In today’s Washington, even the Census Bureau is a source of drama. The department has no director. Due to funding constraints, it has abandoned pre-census research in West Virginia and Washington state that was meant to check the integrity of parts of its survey process. It is weighing whether to add a question about citizenship to the decennial census; community groups around the country have spent months imploring Congress and the Census Bureau not to do so. They’re afraid that adding the question would lower response rates and make the survey less reliable.

At stake: nearly 700 billion federal dollars and how we decide apportion congressional representation.

For groups that work to ensure the census is an accurate count of the population, all those issues are concerns — and ones they didn’t see coming. That’s left less time for the more mundane tasks they had expected to be dealing with at the moment, including one that’s little-known outside census circles1: The census is significantly off in its count of how many young children live in the U.S.

In 2010, some 1 million children under 5 years old didn’t show up in the survey.

And unlike other age groups, for whom the count has improved over time, the count of young kids seems to be getting less accurate. From 1950 to 1980, young kids were about as likely to be counted by the Census as adults were. Since then, however, adults have seen the accuracy of their count improve, and counts of older kids have improved or held steady. Yet more kids under 5 have become invisible to the powerful, constitutionally required survey.

The undercount isn’t a secret; there’s a report on the homepage of the 2010 Census about the problem right now. But a working group gathered by the Census Bureau to examine the problem found that the people responsible for improving the quality of the survey were largely unaware of the issue.

FiveThirtyEight reached out to the Census Bureau for comment, but it did not make someone available before publication.

Debbie Griffin, who retired in the fall of 2017 after working at the Census for nearly four decades, says the organization finally has a much better idea of how bad the problem is. But it’s still missing a lot of information about why. She is hopeful that research questions will be embedded into the 2020 census to help figure out what’s causing the gap, but she worries that the bureau may not have the time or resources to improve on the undercount in the upcoming census.

Those who might pressure the bureau to work on the problem are finding that their energy is demanded elsewhere. “Right now we’re busy fighting the new issue of the citizenship question, which would be an absolute disaster,” said Julie Dowling, a sociologist at the University of Illinois at Urbana-Champaign who serves on the working group convened by the bureau and who wrote a book on how Mexican-Americans think about questions of racial labeling. She’s concerned that adding the question will make data on immigrants less reliable, and fears its impact could be felt most severely in counts of the children of immigrants, the majority of whom are U.S. citizens. And she has one more worry: that the debate over the citizenship question is distracting from the work community groups around the country are supposed to be doing to tackle the undercount of young kids.

“I’m very nervous right now,” said Dowling, “And after investing so much over the years in trying to get an accurate count.”

How it happens

Since the census is the ultimate measure of population in the U.S., one might wonder how we could even know if its count was off. In other words, who recounts the count?

Well, the Census Bureau itself, but using a different data source. After each modern census, the bureau carries out research to gauge the accuracy of the most recent count and to improve the survey for the next time around.

The best method2 for determining the scope of the undercount is refreshingly simple: The bureau compares the total number of recorded births and deaths for people of each birth year, then adds in an estimate of net international migration and … that’s it.3 With that number, the bureau can vet the census — which missed 4.6 percent of kids under 5 in 2010, according to this check.

It’s not exactly clear why so many kids don’t get counted. Outreach costs money, of course, and the Census is experiencing significant constraints. But there are more specific issues, as well.

Some of the undercount arises when kids don’t get included in surveys that are returned, as opposed to living in households for which no census survey is turned in. That may be due to the growing likelihood that young kids will live in complex family situations or with a grandparent. If a kid splits time between two parents who don’t live together, for example, which household counts the child as a member? Research has shown that in households where kids lived with people other than or in addition to their parents — such as multigenerational households — it was more likely that a kid would be left off the census entirely.

The other issue, which may be harder to solve, is the kids who aren’t counted because their entire households are missed. These kids may live in families that are considered hard to count; they live in high-poverty neighborhoods, rental housing or with another family, for example. William O’Hare, a demographer who has advised the census and produced several reports on missed populations, has found that the rate of uncounted young kids in urban areas is double or triple the national figure, and that kids of color are also more likely to be missed.4

The census overlooks more nonwhite children

Percent difference in number of kids aged 0-4 counted in the 2010 census survey and the Census Bureau’s estimate of total kids that age, by race

Race/Ethnic category Percent difference
Black alone -4.4%
Black alone or in combination with other race -6.3
Hispanic -7.5
Non-black, non-Hispanic -2.7

Source: William O’Hare

Some Latino groups have had to prioritize which concern to focus on — the undercount that disproportionately affects kids in their communities, or a citizenship question that could alienate Latinos of all ages from the census. “With the change in administration, the other challenges have been greater,” said Arturo Vargas, the executive director of the NALEO Educational Fund, a nonprofit that promotes Latino participation in the political process (he also serves on the census’s working group). In particular, he’s worried about young Hispanic kids, who are the most likely to be uncounted among young children. Vargas worries that fears among immigrant communities could combine with the existing undercount to form a perfect storm.

Why it matters

The undercount isn’t just a matter of trivia — it has a significant effect on how much funding communities receive from the federal government, and how they are represented by local governments and in Congress. Due to limitations with government data on spending, it’s hard to precisely trace that money trail, especially one tied to a particular age group, but a new study from George Washington University offers a sense of the scale of what’s at stake. The reimbursement formula for Medicaid, the Children’s Health Insurance Program and a handful of other programs5 is directly tied to the population reported in the decennial census. To simulate the effects of a countrywide undercount affecting all age groups, GWU researchers took the population numbers from the 2010 census and reduced them by 1 percent in each state. In the 37 states that qualified for more than the minimum level of federal dollars for these programs, this 1 percent drop in population count would have cost those states a median of $1,091 in 2015 for each person missed by the census, according to the GWU research. Some states would lose out on hundreds of millions of dollars in federal money.

Social programs have a lot of money riding on the census

Estimated federal dollars that welfare programs* would have lost in 2015 if the census has undercounted each state by 1 percent in 2010

* Including Medicaid, the Children’s Health Insurance Program, foster care, adoption assistance, and the Child Care and Development Fund. States with higher per capita incomes already receive the minimum reimbursement rate and would have not lost any funds as a result of an additional undercount.

Source: George Washington University

Collectively, five programs that rely heavily on the census when allocating funds distribute nearly 50 percent of all of the federal government’s grants to the states and represent 13 percent of the value of all state budgets, according to the study’s author, Andrew Reamer, a research professor at GWU. If kids are left out of the total population count, states could get less reimbursement than they should for federally funded programs.

“The irony,” said Vargas, “is that many of the programs that are apportioned by the decennial census are designed to help the same people who are likely to be undercounted.”

That’s a problem for programs like Head Start, which provides early education for children in low-income families. The local programs rely on census data to figure out where they are needed, said Tommy Sheridan, senior director of government affairs at the National Head Start Association. Problems with the big decennial survey can linger for a decade. That dynamic isn’t isolated to Head Start; the census is the bureaucratic bones supporting much of the infrastructure behind our social safety net, political representation and research programs.

It’s not clear what the Census Bureau has planned to improve the count of young kids, said O’Hare, the demographer. That’s concerning, because now the bureau is up against even more challenges than it faced in previous years. “I’ve been watching the census from 1970, and this is clearly the most difficult census I’ve been associated with,” said O’Hare. “A budget squeeze, a leadership vacuum, additional concerns about race and citizenship that some people say sabotages the census. From my point of view, with my interest in [getting an accurate count of] young children, all these problems have made it more difficult to improve on what happened in 2010.”